The Riksbank s inflation target target variable and interval

Transcription

1 The Riksbank s inflation target target variable and interval Riksbank Studies, September 2016 s v e r i g e s r i k s b a n k

2 Production: Sveriges Riksbank Stockholm September 2016 ISBN

3 Riksbank studies Contents Foreword 4 Summary 5 1 Wording of the inflation target The CPI target variable has certain disadvantages Is an interval around the target needed again? 7 2 Which inflation index should the target refer to? Key differences between the CPI, CPIF and HICP Practical aspects that can be relevant to the choice of target index Arguments for and against the CPI, CPIF and HICP 12 3 Should an interval around the target be reintroduced? Different types of targets and intervals Pros and cons of a tolerance band Pros and cons of a target range 15 4 Conclusion 16 References 17 Annex 1 Choice of target index. What does economic theory say? 18 Annex 2 Price indices and housing costs 24 Annex 3 Long-term differences between the rate of increase in the CPI, CPIF and HICP 32

4 4 The Riksbank s inflation target target variable and interval Foreword The Riksbank s inflation target was introduced more than 20 years ago. Since inflation targeting was introduced in 1995, the Riksbank has defined it in terms of the annual change in the consumer price index (CPI). Up until 2010, the Riksbank also used a tolerance band around the inflation target. From the very start, but especially more recently, it has been discussed whether the CPI is the most appropriate target variable. The reason for this is that changes in the repo rate via household mortgage rates have a direct effect on inflation which has nothing to do with the underlying inflationary pressure. The effect is in the opposite direction in that rate cuts aimed at pushing up inflation instead lead to a further fall in CPI inflation in the near term. This makes monetary policy communication more difficult and may lead to inflation expectations being affected in a negative way. Due to this direct interest rate effect, monetary policy decisions have, in practice, been guided by the measure CPIF (CPI with a fixed interest rate), which does not have this effect. Following Goodfriend and King s review of monetary policy last spring, the Riksdag (the Swedish parliament) in June urged the Government to appoint a parliamentary commission of inquiry into the Swedish monetary policy framework and the Sveriges Riksbank Act. 1 The target variable will probably be included in this inquiry. In conjunction with a discussion of the target variable, there may also be reason to consider whether a tolerance band around the target should be reintroduced. The choice of target variable and a possible tolerance band are issues of significant socioeconomic interest. This Riksbank study is to be seen as a basis for a broad and open discussion of these issues. The study discusses the advantages and disadvantages of various possible target variables and of a tolerance band around the target. Changing target variable to CPIF or HICP or reintroducing a tolerance band could have consequences for how the Riksbank communicates monetary policy and builds confidence in the inflation target, but it would not change the basic features of the monetary policy being implemented. In the following discussion, three questions are particularly important: Is there a price index that is preferable to other options or an index that has greater disadvantages? What possible problems are there with a change of target variable for the inflation target? What are the possible advantages and disadvantages of reintroducing an interval around the inflation target? Would a tolerance band or target range be preferable? The Riksbank s role as an independent authority means that the members of the Executive Board may not seek or receive instructions when carrying out their monetary policy tasks (known as the prohibition against instructions). These tasks include specifying the price stability objective introduced into the Sveriges Riksbank Act in On the other hand, both the members of the Executive Board and the Riksbank s experts may engage in discussions with external authorities and organisations on how the target should be specified to create an appropriate and comprehensible monetary policy. The Executive Board of the Riksbank 1 See Evaluation of the Riksbank s monetary policy , Committee on Finance Report 2015/16:FiU41.

5 Riksbank studies Summary Since inflation targeting was introduced in 1995, the Riksbank has defined the target in terms of the consumer price index (CPI). It is now being discussed whether the inflation target should instead be expressed using a different index than the CPI and whether an interval around the target should be reintroduced. 1 One of the reasons is that the CPI is somewhat problematic as a target variable as changes to the policy rate have major direct effects on the CPI, pushing it in the wrong direction. This creates difficulties in the communication of monetary policy and may make target fulfilment difficult if expectations of future inflation are affected. The aim of the inflation-targeting policy is to anchor the long-term inflation expectations of households and companies. This suggests that the inflation target should refer to a relevant, broad and well-known index. The broad CPI with a fixed interest rate (CPIF) measure and the EU-harmonised index for consumer prices (HICP) are the most natural alternatives to the CPI as target variables, if a change is to be made. Present monetary policy is already based on the CPIF, which, like HICP, does not include the direct effects of policy rate adjustments. The CPIF and HICP differ in their coverage and calculation methods, but the outcomes are very similar in practice. In the long term, the differences between the CPIF and the HICP will most probably be small. In the shorter term, over the next ten years, the annual rate of increase in the CPIF is expected to marginally exceed the rate of increase in the HICP. One argument for switching to the CPIF as target variable is that it is probably better known in Sweden than HICP, while international comparability might speak in favour of HICP. The practical consequences for monetary policy of a switch to the CPIF or HICP can be expected to be minor. Regardless of whether a change of target index takes place, the Riksbank will also have to continue to monitor and analyse several measures to obtain the best possible understanding of inflation. A tolerance band is a way of illustrating uncertainty and showing that the Riksbank is unable to fine-tune inflation around the target. It can also provide a concrete expression of the variation in inflation that can reasonably be expected over time. Historical outcomes indicate that a tolerance band has to be very broad if CPI inflation, with reasonable probability, is to be expected to fall within it. With the CPIF or HICP, the interval can be narrower. A target range means that the inflation target itself is defined in terms of an interval, and not just a specific value, such as the current point target of 2 per cent. It therefore provides greater scope for choosing the level of inflation which monetary policy is to aim for. At the same time, this can make it more difficult to anchor long-term inflation expectations than it would be with a point target. Reintroducing a tolerance band could facilitate the communication of monetary policy, but would have limited consequences on the monetary policy 1 See, for example, Jansson (2015).

6 6 The Riksbank s inflation target target variable and interval being conducted. Introducing a target range could have more widespread consequences, as the freedom allowed monetary policy could expand so far as to allow the Riksbank to aim for any inflation rate at all within the interval. In this Riksbank study, we discuss some aspects regarding the choice of target variable for the inflation target and the interval around the target. 1 Wording of the inflation target On 15 January 1993, the General Council of the Riksbank announced that monetary policy was to be guided by an explicit inflation target from then on. The target was specified as the annual change in the consumer price index, the CPI, as from 1995, being limited to 2 per cent, with a tolerance band of ±1 percentage points. 2 When the new Sveriges Riksbank Act came into force in 1999 and an independent Executive Board was tasked with deciding on monetary policy issues, the original wording of the target was retained. 3 For the entire period, the inflation target has been expressed in terms of the CPI. The interval around the target was removed in 2010, however. 1.1 The CPI target variable has certain disadvantages From the very beginning, it was clear that the choice of the CPI as target variable would pose a number of challenges. One reason for this is that changes in the Riksbank s policy rate have direct short-term effects on inflation. When, for example, the Riksbank cuts the repo rate to buoy inflation, mortgage rates will decline. Mortgage rates are included in the owner-occupied housing costs item in the CPI and the rate cut will therefore exert downward pressure on the index. This direct effect on the CPI thus goes in the wrong direction ; in other words, the policy rate cut leads to a drop in near-term inflation. For this and other reasons, the Riksbank has regularly allowed the progression of price indices other than the CPI to influence interest rate decisions, recently mainly the CPIF (CPI with a fixed interest rate). CPIF inflation has served as an intermediate target variable insofar as the Riksbank has usually aimed for CPIF inflation being close to 2 per cent within around two years. The idea is that even if rates sometimes go up and down, CPI and CPIF inflation will eventually coincide when the interest rate stabilises. However, in the past few years, the deviations between the CPI and CPIF have been substantial both upwards and downwards. For example, the period of rate cuts that commenced in December 2011 has contributed to CPI inflation being on average 0.6 percentage points below CPIF inflation. Conversely, CPI inflation will rise quickly and exceed CPIF inflation when the repo rate is raised in the period ahead. Large and protracted differences between CPI and CPIF inflation can cause a number of problems. Perhaps the most serious one is that participants in the economy might start to question whether the Riksbank is actually stabilising CPI inflation around the target by stabilising CPIF inflation around 2 per cent. This can lead to confidence in the inflation target decreasing, and long-term inflation expectations starting to drift away. Another problem is that evaluations of monetary policy are made more difficult if there are major differences between the progression of the variable that guides the Riksbank and the formal target variable. A third problem is that both domestic and overseas analysts often do not take how the CPI is constructed into account when describing the situation in Sweden. This can, for 2 Press release no. 5: Heikensten (1999).

7 Riksbank studies instance, lead to misleading international comparisons. For example, in the past few years media reports have emerged from time to time describing how Sweden is in a state of deflation. This can give the impression of the situation in Sweden being worse, and inflation much lower than in other countries, despite the main difference being that the Swedish CPI is more sensitive to changes in the policy rate than corresponding price indices in other countries. Ultimately, this could also impact inflation expectations. Even if this does not necessarily mean that the CPI has to be replaced as the target variable for monetary policy, there is reason to deepen the discussion on alternative target variables Is an interval around the target needed again? The interval of ±1 percentage points, which complemented the inflation target when it was introduced in 1995, was primarily a pedagogical tool intended to illustrate that certain deviations from the inflation target of 2 per cent, while not being too large, had to be accepted. The size of the interval was determined by what intuitively appeared to be reasonable. It was not possible to rely on any past experience as the monetary policy regime was entirely new and inflation had varied enormously during the old regime which had a fixed exchange rate. It was estimated that inflation would now vary less, but it was unknown as to how much. The interval remained a part of the monetary policy framework for a long time, but was abolished in This was because the Riksbank deemed that it no longer served any practical purpose. In the memorandum published in connection with the decision to abolish the interval, it was noted that: There is considerable understanding for the fact that inflation commonly deviates from the target and that the deviations are sometimes larger than 1 percentage point. Inflation can thus be outside of the tolerance band without threatening the credibility of the inflation target. Such deviations have proved to be a natural part of monetary policy. 5 The question of whether the inflation target should be surrounded by an interval has recently become relevant again in the Swedish debate. Both members of the Executive Board of the Riksbank and others outside the Bank have raised the issue of a possible reintroduction of an interval in some shape or form. 2 Which inflation index should the target refer to? According to the Sveriges Riksbank Act, the objective for monetary policy is to maintain price stability. The Riksbank has specified this as a target for inflation, according to which the annual change in the consumer price index (CPI) is to be 2 per cent. The purpose of inflation targeting is to create stable conditions in the economy, and reduce uncertainty. This makes it easier for households, companies and other participants in the economy to make well-founded economic decisions. In more concrete terms, inflation targeting shall anchor long-term inflation expectations among households and companies. The inflation target can then function as a nominal anchor in the economy and contribute towards stable and effective wage formation. This suggests that the central bank should stabilise a broad and well-known index. Other arguments and more recent monetary policy theory may suggest that the inflation target should refer to a narrower index, such as one where sticky prices carry particularly heavy weight or where an attempt is made to screen out components that monetary policy 4 A change of target variable to CPIF is proposed in Marvin Goodfriend and Mervyn King s review of the Riksbank s monetary policy (Goodfriend and King, 2016). 5 Sveriges Riksbank (2010).

8 8 The Riksbank s inflation target target variable and interval finds difficult to affect (see Annex 1). However, such indices can be difficult to devise. They can also be difficult to communicate because they do not necessarily capture the inflation that is most relevant to households and companies. Therefore, specifying the inflation target in terms of such an index does not currently seem like an alternative that is particularly close to hand. Besides the CPI, there are two broad price indices that are natural alternatives as target variables: The CPI with a fixed interest rate, the CPIF, and the EU-harmonised index for consumer prices, the HICP. Both these are relatively well known. They are based on the same statistical data and have exhibited more or less the same numerical trend. There are, however, certain differences between them. 2.1 Key differences between the CPI, CPIF and HICP Institutional differences The CPI is produced by Statistics Sweden. The calculation rules for the CPI are determined by the Swedish Government and the Riksdag. 6 Until recently, the Consumer Price Index Board (CPI Board) settled issues of a principal nature based on the established grounds. The CPI Board normally had two meetings a year, for which Statistics Sweden drafted proposals for changes in the CPI. 7 The CPI has three principal areas of use. The index is used in calculations of compensation for price developments, 8 in fixed-price calculations 9 and as the target variable for the Riksbank s monetary policy. The CPIF (CPI with fixed interest rate) is produced by Statistics Sweden on behalf of the Riksbank. The only difference between the CPI and the CPIF is that the rate of increase in the CPIF is not directly affected by changes in household mortgage rates. 10 The CPIF has in recent years served as an intermediate target variable for monetary policy insofar as the Riksbank has usually aimed for a CPIF inflation rate of close to 2 per cent within around two years. The HICP is also produced by Statistics Sweden. The calculation rules for the HICP are regulated in EU regulations and other complementary documents on EU level. 11 In many areas, the regulatory framework is made up of minimum rules, where member states may use different methods. The HICP is primarily used as a target value for the ECB s monetary policy, as a comparable inflation measure within the EU and as a basis for evaluations of EU convergence criteria regarding price stability. 12 Differences in calculation method The CPI (and in principle the CPIF) is what is known as a cost-of-living index. A cost-of-living index measures the relationship between the monetary amounts required to maintain, in two price situations, the same consumption standard, or the same level of benefit. 13 This means that situations are compared in which not only the prices but also the consumption composition differ. The index therefore captures the fact that consumers tend to consume more of goods and services that relatively speaking have become cheaper. On average since 1996, this substitution (the basket effect ) has led to CPI inflation being 0.15 percentage points lower per year than it otherwise would have been. 6 The most important calculation rules date from the 1952 Index Committee, the 1955 Housing Index Inquiry and the 1999 Index Inquiry, SOU 1999: New instructions from the Government to Statistics Sweden (2016:822) do not include the stipulation that Statistics Sweden is to be linked to a consumer price index board. This means that the CPI Board will no longer be a decision-making body. 8 In the Government s latest CPI guidelines, Government Bill 2001/02, Annex 4, it is established that the CPI shall primarily form the basis for the most common compensation purposes in society. 9 A fixed-price calculation recalculates nominal values to volume or real values and is used, for example, to analyse households purchasing power and the development of their real incomes. 10 The CPIF is indirectly affected in that the weight for the interest expenditure index can be affected, see Annex 2, which, among other things, summarises how housing costs can be calculated and how they are calculated in the CPIF and the HICP. 11 The foundations are established in Regulation (EU) No 2016/793 of the European Parliament and of the Council. 12 The Bank of England also uses HICP (named CPI in the UK) as a target variable. 13 CPI Inquiry, SOU 1999:124, page 27.

9 Riksbank studies The HICP is a so-called inflation index, i.e. it measures the development in prices of basically the same basket of goods. A key difference between the Swedish CPI and the HICP is therefore that the rate of increase in the HICP is not affected in the same way as the CPI by changes in the consumption pattern. 14 If the development is the same as the historical trend in the period ahead, this difference, in isolation, could therefore lead to CPIF inflation being about 0.15 percentage points lower than HICP inflation. Difference in dealing with owner-occupied housing costs An important difference between the CPI and the HICP is how owner-occupied housing costs are calculated. In the CPI and the CPIF, these expenses are measured as the sum of an estimated capital cost and depreciation as well as day-to-day costs for operation and maintenance (waste collection, insurance, property tax, etc.). 15 The HICP only includes a narrower definition of operating costs. In practice, this means that housing prices are included in the CPI/CPIF, but not in the HICP. However, this occurs with a substantial lag and therefore affects inflation measured in terms of the CPI/CPIF for a long time in the future. As housing prices have increased rapidly for a long time, the contribution of capital costs to CPI/ CPIF inflation will rise in the years ahead, even if housing prices in the period ahead remain unchanged or fall slightly. Since 1996, owner-occupied housing costs have caused CPIF inflation on average to be about 0.15 percentage points higher than HICP inflation. 16 The size of the difference in the period ahead depends, among other things, on how property prices develop. It will also depend on the development of mortgage rates as these affect the weight of the capital cost for owner-occupied housing in the CPI and the CPIF, see Annex 2. For the CPI/CPIF, a new method is currently being investigated where the cost of living in tenant-owned homes is also to be measured using a cost approach. Currently, the costs for living in a tenant-owned home are equated with those for living in a rented home. No decision has been taken, but the aim is for the new method to start to apply from These proposals mean that the difference in CPIF and HICP inflation that stems from housing costs will be slightly greater over the next ten years. This method change was included in the inflation forecast in the July 2016 Monetary Policy Report. Other authorities such as the European statistics agency, Eurostat, are also working on developing an owner-occupied housing item in the HICP. However, this is not calculated the same way as in the CPI (see Annex 3). The issue has been examined and discussed for many years and a decision as to when to start using the item has still not been taken. There are no simple answers for how owner-occupied housing costs should be included in the target index, and it is calculated differently in the CPI, CPIF and HICP. However, in numerical terms, there is no great difference between the CPIF and HICP. Differences from the CPI are more significant as the interest rate has a major effect on capital costs as they are measured in the CPI. 14 Prior to 2005, the Swedish inflation rate was calculated excluding the effect of the updated weights, and reported in parallel with the annual percentage change in the CPI. The Swedish inflation rate was at that time calculated using approximately the same method as in the HICP. In connection with the introduction of a new index construction in 2005, a decision was taken in the Consumer Price Index Board to end this distinction between the inflation rate and the annual percentage change in the CPI. See the CPI Inquiry, SOU 1999:124, Section and the Statistics Sweden memorandum Improved CPI construction from January 2005: Technical description. 15 The capital cost is measured using a so-called Interest expenditure index. This index is affected by how mortgage rates change, and also by how the value of the properties financed by the mortgages changes, in accordance with the following simplified formula: Interest expenditure index = Capital stock index Interest rate index. The capital stock index measures the purchase price of the properties and improvements such as refurbishments and extensions. The interest rate index measures the development of the average mortgage rate. In the CPIF, the interest rate index is kept constant, but the weight for the interest expenditure index can change if mortgage rates change, see Annex 2 for details. 16 Contribution from changes in the capital stock index. 17 Statistics Sweden (2016).

10 10 The Riksbank s inflation target target variable and interval Differences in rate of increase between the CPI, CPIF and HICP As the CPI, CPIF and HICP differ as regards coverage and calculation methods, the rate of increase in the various indices also differs. Falling mortgage rates have pulled down the CPI in relation to both the CPIF and HICP. At the same time, rising property prices and capital costs have caused the CPIF to increase marginally more rapidly than the HICP (see Table 1). 18 Since 1996, the CPI has increased on average by 1.04 per cent per year, while the CPIF and HICP have risen by 1.48 and 1.42 per cent per year respectively. Table 1. Average annual rate of increase in the CPI, CPIF and HICP, January 1996 to March 2016 CPI 1.04 CPIF 1.48 HICP 1.42 Note. Calculations start from 1996 as there were no data for the annual rate of increase in the HICP before then. Source: Statistics Sweden In the period ahead, the differences in rates of increase will depend on the development of mortgage rates and property prices, basket effects and any changes in the calculation methods and their implementation. Based on the calculations presented in Annex 3, we can draw the following conclusions: In the long term, the differences between the CPIF and HICP will be small given reasonable assumptions about the basket effect, growth in property prices and the size of capital costs for owner-occupied housing as a proportion of consumption expenditure. Assuming that the repo rate will stabilise in the long term, the rate of increase in the CPI and CPIF will be the same. Over the next five years, expected repo rate increases will cause the CPI to rise much more rapidly than the CPIF and HICP. At the same time, we can expect the rate of increase in the CPIF to be slightly higher than the rate of increase in the HICP. The high rate of increase in single-family and tenant-owned housing prices in recent years will lead to higher housing costs in the CPIF than in the HICP over the next five years. In the medium term, in about ten years time, the annual rate of increase in the CPIF may, however, exceed the rate of increase in the HICP by around 0.2 percentage points. This is due partly to the fact that the weight for capital costs will rise in the CPIF when the interest rate goes up and to the fact that the substantial rate of increase in the prices for single-family dwellings and tenant-owned homes in recent years will lead to rising housing costs in the CPI/CPIF for a long time to come. The isolated effect of this will be that the CPIF is expected to increase by 0.35 per cent more rapidly than the HICP per year. But the effect is counteracted to a certain extent by the basket effect in the CPIF. Assuming that the basket effect is 0.15 percentage points per year, the rate of increase in the CPIF, according to these calculations, will be 0.2 percentage points higher than in the HICP in total. Once repo rate increases stop and have remained unchanged for several years, the annual rate of increase in the CPI in ten years is expected to be about the same as the rate of increase in the CPIF. If the HICP were to include owner-occupied housing costs using a net acquisition approach in accordance with the proposal currently being analysed by Eurostat it would probably lead to a slightly higher rate of increase in the HICP as construction costs normally increase more rapidly than other consumer prices. 19 The differences in the rate of increase between the CPI/CPIF and the HICP would therefore be smaller. 18 See Annex 3 for details. 19 In a net acquisition approach, the price progression of new houses are monitored. According to the proposal currently being analysed by Eurostat, land prices are to be excluded.

11 Riksbank studies As mentioned above, a discussion is underway on calculating the costs of living in tenantowned homes using a cost approach similar to the one used for owner-occupied housing. In addition, it has been decided that tax relief on interest expenditure is to be included in the CPI as of 2017, so that household interest expenditure is measured after tax. Calculations indicate that the weight for interest expenditure in the CPI and CPIF would increase by just over 40 per cent if tenant-owned homes were included. Including the tax relief on interest expenditure in the CPI simultaneously reduces the interest weight (for owner-occupied and tenant-owned homes) by 30 per cent. Both these method changes are included in the inflation forecast as from the July 2016 Monetary Policy Report. In a long-term equilibrium, when the interest rate is stable for a longer time, the CPI and CPIF will increase at the same rate. But across economic cycles, when the interest rate varies around a certain level, it is possible that the CPI will increase more rapidly than the CPIF, as the percentage change of a rate rise is greater than the percentage change of a rate cut. 20 Given reasonable assumptions, the average rate of increase in the CPI can be about 0.15 percentage points higher than the average rate of increase in the CPIF for long periods, see Annex Practical aspects that can be relevant to the choice of target index If the Riksbank changes the target variable for monetary policy from the CPI to the CPIF or the HICP, Statistics Sweden will continue to produce and publish the CPI. As explained above, a possible consequence of any change in the target variable for monetary policy could, however, be that the CPI increases slightly more rapidly than the new target variable. This can have consequences that are not directly related to monetary policy. Several taxes and public expenditure items have a direct connection to inflation measured as the CPI or indirectly via the price base amount. 21 This applies above all to social protection systems directed at households and certain specific taxes. About a quarter of the Government s total expenditure has a direct connection to the CPI or the price base amount. Any change in the target variable for monetary policy could therefore affect public finances unless the rate of increase in the CPI coincides with the new target variable. If, for example, the target variable for monetary policy were to be changed from the CPI to the CPIF, and the target was still set at 2 per cent, it is probable that the price base amount and the CPI would increase slightly more quickly than 2 per cent in the medium term. The net effect on public finances is difficult to assess, but is probably limited. Financial markets are affected by the CPI mostly as a result of the Swedish National Debt Office issuing government bonds that are indexed to the CPI. 22 If a change in target variable leads to the CPI increasing more rapidly than is currently expected, investors currently holding index-linked bonds will make capital gains. This effect is short-lived, however, and market prices will soon adapt. Another aspect is that the survey on inflation expectations performed by TNS Sifo Prospera on behalf of the Riksbank has been based on the CPI up until now. The survey would therefore need to be supplemented with questions linked to the new measure. Time series breaks in the survey can to a certain extent be bridged over by taking parallel readings of both measures. 20 The percentage change in the average mortgage rate is greater when the interest rate goes up from, say, 6 to 6.5 per cent than when the interest rate falls from 6.5 to 6 per cent (even if the change in percentage points is the same). Over an economic cycle, where the repo rate is raised as much as it is cut, the average percentage change in mortgage rates will therefore be greater than zero. 21 Every year, the Government decides on the price base amount that is used, for example, in the social insurance and tax systems. In 2016, the price base amount was SEK The price base amount is adjusted upwards with the help of the CPI inflation rate in June of the immediately preceding year. As the price base amount is rounded off to the nearest SEK 100, the annual changes in the CPI and the price base amount are not identical. 22 Inflation-indexed bonds are common in Sweden and their nominal yield is equal to the sum of a fixed real interest rate and a part that varies with inflation. The prices of these bonds can be used to estimate implicit inflation expectations.

12 12 The Riksbank s inflation target target variable and interval The CPI is also used in other contexts, for example, for compensation purposes and to adjust prices in various agreements. For instance, the price amount guides the statutory minimum amount when it comes to distribution of an estate, inheritance or will. Insurance pay-outs are also often given in terms of a number of price base amounts. Most commercial rents and many leaseholds are annually recalculated using the CPI and business agreements containing index clauses often express these in CPI terms. The CPI is also used to recalculate nominal amounts to fixed prices and volumes, i.e. deflate. When calculating real wages, it is common to deflate using the CPI. Any change in the target variable should, however, be unproblematic in this context. The inflation target is to work as a benchmark for expectations in the economy and thereby lay the foundations for efficient price-setting and wage formation. Should a change be implemented, both the CPIF and the HICP could work as a good benchmark for expectations. Both are broad indices and capture the inflation that is relevant for consumers and wage earners. It is probable that both the CPIF and HICP could quickly become well-known inflation measures if either of them were to become the target variable for monetary policy. 2.3 Arguments for and against the CPI, CPIF and HICP Any change in target variable may need to take the following pros and cons into consideration: CPI The Riksbank s reason for selecting the CPI as the target index was that it is a broad price index that represents ordinary purchases. Also, CPI statistics are of good quality, are published shortly after the end of the month and are not usually revised. The CPI is a wellknown measure among the Swedish general public, and is used in a number of contexts. These reasons are naturally still valid. Keeping the CPI as the target index signifies continuity. A change in target variable may lead to expectations that it may be replaced again. This can increase uncertainty about future monetary policy. As discussed above, the CPI does have a number of disadvantages, primarily because changes in the Riksbank s policy rate have direct, short-term effects on inflation. CPIF The CPIF shares all the same properties as the CPI, apart from the disadvantage of being directly affected by interest rate adjustments via mortgage rates. One advantage of the CPIF in comparison with the HICP is that the Riksbank already uses the CPIF as a monetary policy rudder, and would continue to do so. The possible disadvantages of the CPIF compared with the HICP are as follows: The CPIF includes housing prices with a significant lag and affects inflation measurements for a long time in the future. As housing prices have increased rapidly for a long time, capital stock will contribute positively to CPIF inflation in the years ahead even if housing prices were to remain unchanged or fall. Monetary policy may therefore react to a rise in housing prices that took place a long time ago. The HICP does not have this problem. The CPIF has been explicitly developed to work as a complement to the CPI. One can therefore claim that it is not a sufficiently independent measure. Statistics Sweden produces the CPIF at the request of the Riksbank. Having the inflation measure calculated by an institution other than the central bank makes it easier to maintain confidence in inflation targeting. However, the fact that the Riksbank itself has designed the index and decided how it should be calculated could be seen as a problem.

13 Riksbank studies HICP Neither is the HICP directly affected by interest rate adjustments as mortgage costs are not included in the measure. An advantage of the HICP in comparison with the CPIF is that it is used as a target index by Sweden s neighbouring central banks (the ECB and the Bank of England). In addition, it is often used for comparisons within the EU. If it is considered that monetary policy should stabilise the increase in a pure price index, rather than in a cost-ofliving index, the HICP is preferable. The possible disadvantages of the CPIF compared with the HICP are as follows: Currently, the HICP completely excludes the capital costs for owner-occupied housing. 23 The HICP does not take into account consumers tendency to go over to cheaper goods and services as much as the CPIF does, which leads to households living costs being overestimated. If it is considered important that the monetary policy target variable measures households living costs, this can be seen as a disadvantage. The HICP is probably not as well-known as the CPI or CPIF. This would in all likelihood change if it became the new target variable for monetary policy in Sweden. 3 Should an interval around the target be reintroduced? The question of whether the inflation target should be surrounded by an interval has once again become relevant in the Swedish debate. To facilitate the discussion, it is first necessary to define a few concepts. 3.1 Different types of targets and intervals A point target is an inflation target where the target is defined in the form of a specific value for the rate of change in a target index, often in annual terms. For example, Sweden has a point target that says that the annual rate of change in the CPI should be 2 per cent. Tolerance band A point target can be supplemented with a tolerance band. The central bank then pursues a monetary policy in which inflation is to reach the point target, but the tolerance band shows which inflation outcomes can be tolerated or are counted as acceptable. If inflation ends up outside the interval if the deviation is greater than what is acceptable some form of sanction may be triggered in certain countries. In New Zealand and the United Kingdom, for instance, the central bank must present to the Government the reasons why inflation has ended up outside the interval when this happens. The interval then forms part of the agreement drawn up between the central bank and its governing body. The tolerance band can also indicate the central bank s own level of ambition with respect to stabilising inflation, without outcomes outside the interval giving rise to any kind of sanction. The interval used by the Riksbank prior to 2010 was of this type. A tolerance band can also illustrate the fact that inflation is continually affected by different shocks and is difficult to control with any great degree of precision. The Riksbank s earlier interval also had this function This will change if/when Eurostat decides to introduce an owner-occupied housing item in the HICP. 24 The stated purpose of the interval was to illustrate that deviations from the inflation target are probable, and that the Riksbank s aim was to try to limit these deviations (see, for example, Heikensten, 1999).

14 14 The Riksbank s inflation target target variable and interval Target range A target range is an inflation target where the actual target is defined in the form of an interval for the percentage change in a target index. With a target range, there is no requirement for inflation to reach the exact midway point in the interval. The central bank can, in principle, pursue a monetary policy in which inflation is stabilised just before the boundary of the interval. Australia is an example of a country where the inflation target is defined as a target range. 25 The differences between a tolerance band and a target range is illustrated in Chart 1. The tolerance band thus refers to outcome and monetary policy is always aimed at bringing inflation onto the point target. The target range is, on the other hand, also forward-looking insofar as monetary policy can be aimed at achieving any point within the interval Chart 1. Tolerance band vs target range Per cent Tolerance band Target range Note. The unbroken line denotes the outcome, i.e. the percentage annual rate of change in the CPIF. The broken blue lines in both left-hand and right-hand charts represent the forecast from the Monetary Policy Report, July The broken red lines in the right-hand chart represent fictitious forecasts. Sources: Statistics Sweden and the Riksbank 3.2 Pros and cons of a tolerance band A tolerance band could facilitate monetary policy communication. It would signal that the Riksbank has the ambition of limiting the variation in inflation and simultaneously provide a concrete benchmark for the variation that could be expected over time. Such an interval could thereby fulfil the same function as the interval that was abolished in The Riksbank already has channels for communicating the fact that there is uncertainty concerning the development of inflation. One such channel is the Riksbank s forecasts. The inflation forecasts show that it can take time during certain periods before inflation gets back to the target and that inflation is not expected to be exactly on target all the time. Furthermore, uncertainty is illustrated in the forecasts with uncertainty bands. It is possible, however, that a tolerance band is a more pedagogical way of illustrating this uncertainty and that economic agents would perceive a tolerance band as a clearer alternative. It could therefore work as a complement to the existing communication. It is also possible that a tolerance band would make deviations from the point target easier to accept and not give rise to criticism as long as inflation stayed within the interval. As such criticism can in itself reduce confidence in the inflation target, it cannot be ruled out that a tolerance band could, as a result, indirectly help to keep long-term inflation 25 In pursuing the goal of medium-term price stability, both the Reserve Bank and the Government agree on the objective of keeping consumer price inflation between 2 and 3 per cent, on average, over the cycle. This formulation allows for the natural short-run variation in inflation over the cycle while preserving a clearly identifiable performance benchmark over time. Reserve Bank of Australia (2013).

15 Riksbank studies expectations better anchored to the target. A disadvantage may, however, be that if inflation does fall outside the interval, it could conversely be perceived as particularly serious. A further complication is that the interval must also be well adapted. One reason for abolishing the previous tolerance band in 2010 was that CPI inflation had been outside the interval just as often as it had been within it. If the interval is to work as a useful benchmark for the variation in inflation that can be reasonably expected, the interval should be realistic insofar as a large proportion of inflation outcomes can be expected to fall within it. The choice of target variable is therefore significant when it comes to setting a suitable interval range. A tolerance band of ±1 percentage points around the inflation target would work better for less volatile inflation measures such as the CPIF and HICP than it would for the CPI. An interval of that size would cover about 70 per cent of the outcomes for the CPIF since 1996 and about 60 per cent of HICP outcomes. To cover about 90 per cent of outcomes, the interval for the CPIF would need to be ±1.5 percentage points and almost ±2 percentage points for the HICP. The exact range of any new interval is therefore an issue that requires careful consideration. At the same time, it is appropriate, not least for pedagogical reasons, to choose a figure that is easy to remember. A well-adapted range for the tolerance band would in itself make any new interval more meaningful than the previous one. There may also be reason to consider whether or not outcomes outside the interval should lead to any particular consequences and, if so, which. At one end of the scale, the interval would only illustrate that the development of inflation is uncertain and outcomes outside the interval would not lead to any particular consequences. At the other end, a possible option would be to oblige the Riksbank, in the same way as the Reserve Bank of New Zealand and the Bank of England, to explain to parliament why inflation had fallen outside the interval. 3.3 Pros and cons of a target range Introducing a target range would be a significantly greater step than introducing a tolerance band and the consequences could be much more sweeping. With a target range, the freedom allowed monetary policy could increase in such a way as to allow the Riksbank to aim for any inflation rate at all within the interval. At the same time, it could also make it more difficult to keep long-term inflation expectations anchored if the target is set without a point target and where all levels of inflation within the interval of, say, 1-3 per cent, are deemed to fulfil the inflation target. The nominal anchor in the economy would then be more unclear. This could make wage formation more difficult, for example. With a target range, it is also important to avoid inflation and inflation expectations permanently being in the lower half of the interval. It would then be more difficult to stimulate the economy in the future when economic activity is weak or inflation is below the target range. This is due, among other things, to the fact that policy rates are on average lower when inflation is on average low. This reduces scope for cutting the policy rate, as this cannot be decreased indefinitely and will hit its lower bound more frequently. When average inflation is low, it is more difficult to achieve the really low or even negative real interest rates that are sometimes needed to stimulate the economy. In the international debate following the financial crisis, proposals have therefore been put forward to raise the inflation targets of central banks, simply to increase monetary policy s scope for action. 26 One way of both preventing inflation from being permanently close to its lower bound and reducing the risk of inflation falling outside the target range is to aim for the midway point of the interval. But then the target range basically becomes a point target. 26 See, for example Blanchard et al. (2010).

16 16 The Riksbank s inflation target target variable and interval 4 Conclusion In this Riksbank study, we have discussed various aspects regarding the choice of target variable for the inflation target and the interval around the target. No measure is entirely problem-free the CPI, CPIF and HICP all have their advantages and disadvantages. In recent years, monetary policy has been based, in practice, on the CPIF which has tended to develop in a similar manner to the HICP, in numerical terms. Changing target variable to CPIF or HICP or reintroducing a tolerance band could have consequences for the Riksbank s communication of monetary policy and its ability to build confidence for the inflation target. However, the basic features of the monetary policy being conducted would not be affected appreciably. Our hope is that this Riksbank Study will contribute towards an increased understanding and discussion of the relevant issues.

17 Riksbank studies References Blanchard, Olivier, Dell Ariccia, Giovanni and Mauro, Paolo (2010), Rethinking Macro Policy. IMF Staff Position Note SPN/10/03. International Monetary Fund. Goodfriend, Marvin and King, Mervyn (2016), Review of the Riksbank s monetary policy Report of the Riksdag Committee on Finance 2015/16:FiU41. Heikensten, Lars (1999), The Riksbank s inflation target clarifications and evaluation. Sveriges Riksbank Economic Review 1999:1. Sveriges Riksbank. Jansson, Per (2015), Time to improve the inflation target? Speech at Handelsbanken, Stockholm, 3 December. Reserve Bank of Australia (2013), Statement on the Conduct of Monetary Policy, The Treasurer and the Governor of the Reserve Bank, 24 October 2013, SCB (2016), Beslut om förändrad hantering av ränteavdraget i KPI [Decision to change how tax relief on interest expenditure is dealt with in the CPI]. Press release, 23 May 2016, sv_/ Om-SCB/Nyheter-och-pressmeddelanden/Behallare-for-Nyheter-och-Pressmeddelanden/Beslutomforandrad-hantering-av-ranteavdraget-i-KPI/#. Sveriges Riksbank (2010), The Riksbank removes the tolerance interval from its specified monetary policy target. Memorandum, Basis for Decision, 31 May 2010, Dokument_riksbank/Kat_publicerat/Pressmeddelanden/2010/nr27e_beslutsunderlag.pdf.

18 18 The Riksbank s inflation target target variable and interval Annex 1 Choice of target index. What does economic theory say? This annex provides a brief summary of the more theoretical aspects regarding the choice of target index for the inflation target. 1 Opinions are divided as to the characteristics a target index should have. The opinion that continues to dominate in practice is that the central bank is to stabilise a broad index with the aim of anchoring inflation expectations in the economy. Traditional view stabilising a broad index and anchoring inflation expectations The purpose of inflation targeting is to create stable conditions in the economy, and reduce uncertainty. This makes it easier for households and companies to make well-founded financial decisions. In more concrete terms, inflation targeting should anchor the long-term inflation expectations of households and companies the aim being to give the economy a nominal anchor. A nominal anchor is important because it is otherwise easy to end up in a price and wage spiral, in which economic policy makers find themselves obliged to conduct a policy that fulfils high inflation expectations a self-perpetuating process. Although such a spiral does not last infinitely, inflation often ends up at an undesirable high level. The development in Sweden in the decades preceding the introduction of the inflation target in 1993 is an example of this. The economy was then stuck in a devaluation cycle, in which recurring cost crises due to excessive wage increases were addressed by writing down the value of the Swedish krona. When the inflation target was introduced, the primary problem was that inflation expectations were too high. It is, however, equally important to prevent expectations from starting to drift downwards, because a period of deflation can also have negative implications for the real economy. What does this say about the choice of target index? If the inflation target is to constitute a nominal anchor, it says that the target should refer to a price index that is known and relevant to those who set wages and prices. If the central bank stabilises a price index that is not perceived as relevant, inflation expectations can start to deviate, even if the anchor remains in place. This suggests that the central bank should stabilise a broad and wellknown index, such as the CPI. 2 New view stabilising sticky prices The past few decades have seen the emergence of a new monetary policy theory New Keynesian theory with a partly different view of which target variables are appropriate. In this theory, macroeconomic relationships are built up from microeconomic theory about the behaviour of households and companies. Economic models devised according to this theory 1 This annex is based on Apel, Armelius and Claussen (2016). 2 Even before inflation targeting became commonplace, it was argued that monetary policy should aim for broad indices, see, for example, Wynne (2008). There is also an older discussion indicating that asset prices should be included in the index, see, for example, Alchian and Klein (1973)) and Goodhart (2001).